I help startups ship AI systems that move from demo to production: MCP tools, fine-tuning pipelines, inference workflows, CV automation, and full-stack product surfaces.
Published
Top Rated100% Job SuccessProjects
AI infrastructure project
MCP-native fine-tuning workflow that moves from natural-language model intent to packaged model artifacts.
Fine-tuning required manual dataset prep, training scripts, GPU setup, eval tracking, packaging, and reporting.
Worked on building an MCP-native fine-tuning control plane using Python, FastAPI, MCP tools, YOLO, and ONNX to turn natural-language training requests into documented model artifacts.
Production platform work around a self-hosted, user-owned AI system with modular plugins and hosted services.
A self-hosted AI platform needed reliable product surfaces for model access, hosted workflows, usage paths, and release operations.
Worked on building production AI platform surfaces using React, TypeScript, FastAPI, Python, and hosted model workflows to improve self-hosted and hosted AI product infrastructure.
Computer vision demo
Real-time sports analytics prototype with player IDs, pose, ball tracking, and stable visual overlays.
Sports footage is messy: occlusion, small ball motion, player overlap, and camera movement make reliable overlays hard.
Worked on building a sports computer-vision pipeline using YOLOv8, TrackNet, DeepSORT, MMPose, and OpenCV to produce player tracking, ball tracking, pose cues, and stable demo overlays.
Experience
AI infrastructure project
AI/ML Solutions · B2B Consulting
Core AI Engineer
Demos
Computer vision demo
Sports CV pipeline for tracking players, pose, ball movement, and stable visual overlays from match footage.
LLM evaluation support
Supporting BrainDrive work for comparing models on product tasks like coverage, factuality, and hallucination risk.
Client CV/OCR build
Real-time mobile OCR for One Piece TCG cards, tuned for motion blur, perspective, and tiny ID regions.
Research
Published work on micro-expression recognition using Swin-T on CASME II, comparing spatial and temporal heads, with gains from temporal cues and RAF-DB pretraining.
GitHub
Loading contributions…
Education
